A school professional knowledge graph construction method and system

Abstract

The present application belongs to the field of information matching and analysis, and particularly relates to a school professional knowledge graph construction method and system. The school professional knowledge graph constructed by the present application not only distinguishes the similarity by professional names, but also calculates more accurate similarity by combining multi-dimension and multi-method, and can solve the problems of talent omission and recruitment and employment cost increase encountered by students in the employment process and recruiters in the recruitment process. Furthermore, the school professional knowledge graph constructed not only contains professional association relationship, which can enable college job seekers and enterprise recruiters to more clearly understand the association relationship between the major and the required major, but also contains the information association relationship of each school major itself, which can guide the job-seeking students to deliver the work with the same direction but different professional names, and can also help the enterprise recruiters to preliminarily evaluate the actual ability of the job seekers from the school major aspect, so as to facilitate the judgment of whether the job seekers can be competent for the corresponding work.

Description

Technical Field

[0001] This invention belongs to the field of information matching and analysis, specifically relating to a method and system for constructing a school knowledge graph. Background Technology

[0002] The current method of matching graduate employment and corporate recruitment needs usually involves corporate recruiters setting the professional requirements for each position in the job postings. Job seekers then select positions that match their majors based on the professional requirements in the job postings and submit their resumes. After receiving the resumes, corporate recruiters will also judge whether the applicants' majors meet the requirements of the corresponding positions based on the school majors in their resumes.

[0003] However, with the popularization of education and the expansion of university enrollment, the number of college students is increasing. At the same time, the rapid development of technology has also spawned a large number of new industries. In order to adapt to these new industries, many new majors have been added. Moreover, there are some differences between the names of majors in different schools and the national standard major names, as well as between majors in different schools. When recruiters post job openings, they can only set the professional requirements for the positions based on their experience. There is no unified standard for the names of majors. As a result, it is difficult for job seekers to determine whether their major matches the job requirements when applying for a position. It is also difficult for recruiters to determine whether job seekers' majors are suitable for the positions when reviewing resumes. Ultimately, this may lead to the loss of talent and an increase in recruitment and employment costs. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing a school professional knowledge map, which solves the problem that the lack of a unified standard for information matching of professional names in the existing technology makes it difficult for both job seekers and employers to determine whether a job seeker's school major matches the job requirements, resulting in talent omissions and increased recruitment and employment costs.

[0005] To achieve the above objectives, this invention provides a method for constructing a school knowledge map, comprising the following steps:

[0006] 1) Obtain data related to schools and majors; the data related to schools and majors includes national standard major set data, major set data of each school, and course data of each school's majors; the majors of each school include different majors offered by different schools and different majors offered by the same school.

[0007] 2) Preprocess the data related to the schools and majors to obtain a set of words corresponding to the schools and majors;

[0008] 3) Based on the set of words corresponding to schools and majors, obtain the professional similarity between majors and the professional association between majors accordingly; the professional similarity between majors includes the professional similarity between each national standard major and each school major, as well as the professional similarity between majors in each school.

[0009] 4) Based on the acquired school and major-related data, obtain the information relationships between each school and major; based on the professional relationships between the majors and the information relationships between each school and major, generate a school professional knowledge map.

[0010] The beneficial effects of the above technical solution are as follows: the constructed school professional knowledge graph can combine multi-dimensional school and major-related data to obtain a more realistic professional similarity, which makes the judgment on whether the job seeker's school major matches the job's professional requirements more accurate, thereby better solving the problems of talent omission and increased recruitment and employment costs encountered by college students in the process of employment and corporate recruiters in the recruitment process.

[0011] Furthermore, the method for preprocessing the school and major-related data to obtain the word set corresponding to the school and major is as follows:

[0012] A word segmentation model is constructed, a stop word dictionary and a word segmentation dictionary are loaded, and each piece of data in the acquired school and major-related data is processed by the word segmentation model to remove stop words; the stop words include modal particles and / or meaningless words;

[0013] Each data point after removing stop words is segmented using a word segmentation model to form various word sets. Each word set corresponds to a data point related to a school and its major, and the word sets do not affect each other.

[0014] Furthermore, based on the sets of words corresponding to schools and majors, the method for obtaining the professional similarity between majors is as follows:

[0015] Based on the word sets corresponding to schools and majors, the name similarity, semantic similarity, and course similarity among majors are calculated respectively; the name similarity, semantic similarity, and course similarity among majors include the name similarity and semantic similarity between each national standard major and each school major, as well as the name similarity, semantic similarity, and course similarity between majors among schools.

[0016] The similarity scores of the names and semantics of each national standard major compared with each school's major are multiplied by their respective weighting coefficients and then summed to obtain the major similarity scores of each national standard major compared with each school's major. The similarity scores of the names, semantics, and courses of each school's majors are multiplied by their respective weighting coefficients and then summed to obtain the major similarity scores of each school's majors.

[0017] The beneficial effects of the above technical solution are as follows: by weighting and summing the similarities in different aspects to obtain the professional similarity between various majors, the obtained professional similarity can more accurately and comprehensively reflect the actual degree of similarity between various majors and improve the accuracy of professional similarity calculation.

[0018] Furthermore, the term sets corresponding to schools and majors include term sets corresponding to national standard major set data and term sets corresponding to each school's major set data; wherein, each term set corresponding to a national standard major corresponds to a national standard major name, and each term set corresponding to a school's major corresponds to a school's major name;

[0019] The method for calculating the name similarity between each national standard specialty and each school specialty, and between different school specialties, is as follows:

[0020] The words in the word sets corresponding to the national standard professional set data and the professional set data of each school are merged, deduplicated, and sorted to form a professional corpus dictionary corresponding to the majors;

[0021] Based on the professional corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the national standard major and each word set corresponding to the majors of each school: For a word set, several vectors are generated, each with a component of 0 in each dimension. Each vector corresponds to a word in the word set, and the dimension of these vectors is determined by the length of the professional corpus dictionary; then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, thus obtaining the position vector corresponding to each word in the word set;

[0022] The position vectors corresponding to all words in each word set are merged in dimension to form vectors that contain the position information of all words in each word set corresponding to the national standard major and the majors of each school in the professional corpus dictionary. Then, a similarity algorithm is used to calculate the similarity between each merged vector and each school major. The calculated similarity is used as the name similarity between each national standard major and each school major or the name similarity between each school major.

[0023] Furthermore, the semantic similarity calculation method for comparing each national standard specialty with each school specialty, and for comparing the semantic similarity between school specialties, is as follows:

[0024] The similarity of names between each national standard major and each school major, or between different school majors, is assessed. If the name similarity between each national standard major and each school major, or between different school majors, is greater than or equal to the name similarity threshold, the corresponding semantic similarity is directly set to the first preset value. Otherwise, a word similarity calculation method based on a thesaurus is used to calculate the semantic similarity for each word set corresponding to the national standard major and each word set corresponding to each school major. The calculated similarity is used as the semantic similarity for each national standard major compared to each school major, and for each school major compared to each other.

[0025] The beneficial effects of the above technical solution are: when name similarity judgment may fail, semantic similarity can be used as a supplement, which can minimize the impact of professional similarity calculation on the multiple variations of professional names.

[0026] Furthermore, the term set corresponding to schools and majors also includes the term set corresponding to the course data of each school's majors; among them, the term set corresponding to the course data of each school's majors corresponds to all courses offered under the name of a school's major.

[0027] The method for calculating the course similarity between the various school majors is as follows:

[0028] The word sets corresponding to the course data of each school's majors are merged, deduplicated, and sorted to form a course corpus dictionary corresponding to the courses offered by each school's majors.

[0029] Based on the course corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the course data offered by each school major: several vectors with each dimension component set to 0 are generated for each word set, each vector corresponding to each word in the word set, and the dimension of these vectors is determined by the length of the course corpus dictionary; then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, thus obtaining the position vector corresponding to each word in the word set;

[0030] The position vectors corresponding to all words in each word set are merged in dimension to form vectors that contain the position information of all words in each word set in the course corpus dictionary. Then, a similarity algorithm is used to calculate the similarity between each merged vector, and the calculated similarity is used as the course similarity between each major.

[0031] Furthermore, the corresponding method for obtaining the professional relationships between various majors is as follows:

[0032] The similarity between each national standard major and each school major, as well as between majors in each school, is compared with the set similarity threshold and equality threshold, where the similarity threshold is less than the equality threshold.

[0033] If two majors have a similarity greater than or equal to the similarity threshold but less than the equality threshold, then the major association between the two majors is set to similar; if two majors have a similarity greater than or equal to the equality threshold, then the major association between the two majors is set to equal; if two majors have a similarity less than the similarity threshold, then it is determined that there is no major association between the two majors, and the corresponding data is discarded.

[0034] The beneficial effects of the above technical solution are as follows: determining the professional relationship based on professional similarity and discarding some data accordingly can reduce redundant data and avoid the impact of professional-related data without professional relationship on the display of professional relationships, resulting in the final knowledge graph being less intuitive.

[0035] Furthermore, the data related to schools and majors also includes basic data on each school's majors and relevant information about each school; the basic data on each school's majors includes major ranking, major training direction, and data on former names of the major; the relevant information about each school includes school ranking and school level data.

[0036] The information relationships between each school's majors include the relationships between basic information about the school's majors, the relationships between basic information about the school, and the relationships between the courses offered by the school's majors.

[0037] Based on the obtained data related to schools and majors, the information relationships between each school and its majors are obtained in the following ways: by obtaining the basic information relationships between schools and majors through the basic data of each school's majors; by obtaining the basic information relationships between schools through relevant information of each school; and by obtaining the information relationships between the courses offered by each school's majors through the data of the courses offered by each school's majors.

[0038] The beneficial effects of the above technical solution are as follows: the constructed school professional knowledge map not only includes professional relationships, enabling college students seeking employment and corporate recruiters to more clearly understand the relationship between their major and the required major, but also includes the information relationships of each school's major itself. It can guide job seekers to apply for jobs with different major names but the same direction, and can also help corporate recruiters to conduct a preliminary assessment of job seekers' actual abilities from the perspective of school majors, making it easier to judge whether job seekers are competent for the corresponding jobs.

[0039] Furthermore, based on the aforementioned professional relationships and the information relationships within each school's respective majors, the method for generating a school-specific knowledge graph is as follows:

[0040] Based on the acquired school and major-related data, different entities are constructed; the constructed entities include national standard major entities, school major entities, school name entities, school basic information entities, school major basic information entities, and major-offered course entities.

[0041] The constructed entities are matched with the professional relationships between various majors and the information relationships between the majors of each school, respectively, to generate a school knowledge knowledge graph.

[0042] The present invention also provides a system for constructing a school knowledge graph, including a processor, the processor being used to execute program instructions to implement the above-described method for constructing a school knowledge graph.

[0043] The system for constructing the school's professional knowledge graph can achieve the same beneficial effects as the aforementioned method for constructing a school's professional knowledge graph. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the overall construction method of the school professional knowledge graph in an embodiment of the construction method of the present invention;

[0045] Figure 2 This is a schematic diagram of school and major-related data obtained in an embodiment of the method for constructing a school professional knowledge graph according to the present invention;

[0046] Figure 3 This is a flowchart illustrating the preprocessing of school and major-related data in an embodiment of the method for constructing a school professional knowledge graph according to the present invention.

[0047] Figure 4 This is a flowchart illustrating the process of one-hot encoding for each word in each set of words corresponding to national standard majors and each set of words corresponding to majors in each school, as well as in an embodiment of the method for constructing a school professional knowledge graph according to the present invention.

[0048] Figure 5 This is a flowchart illustrating the calculation method for semantic similarity between different majors in an embodiment of the school professional knowledge graph construction method of the present invention.

[0049] Figure 6 This is a flowchart illustrating the calculation method for course similarity among different majors in an embodiment of the school professional knowledge graph construction method of the present invention.

[0050] Figure 7This is a flowchart illustrating the process of calculating the similarity between different majors in an embodiment of the method for constructing a school professional knowledge graph according to the present invention.

[0051] Figure 8 This is a schematic diagram illustrating the method for generating a school knowledge graph in an embodiment of the school knowledge graph construction method of the present invention;

[0052] Figure 9 This is a schematic diagram of a school knowledge graph generated in an embodiment of the method for constructing a school knowledge graph according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0054] Example of a method for constructing a school's professional knowledge graph:

[0055] This embodiment provides a technical solution for constructing a school professional knowledge map, referring to... Figure 1 ( Figure 1 The construction method (which represents each step as a corresponding processing module) has the following specific steps:

[0056] 1) Obtain relevant data on schools and majors; refer to Figure 2 The data acquired through the data collection module mainly includes national standard professional set data (e.g., "Professor A, Professional B, Professional C..."), professional set data of each school (e.g., "A school's professional set includes Professional A, Professional B, Professional C..."), and course data offered by each school's professional set (e.g., "A school's professional set includes Course A, Course B, Course C..."). It also includes basic data on each school's professional set and related information. Basic data on each school's professional set includes professional ranking, professional training direction, and data on the professional set's former names. Related information includes school ranking and school level. To ensure comprehensiveness, the professional set data includes different professional sets offered by different schools and different professional sets offered by the same school.

[0057] It should be clarified that the sources of school and major-related data are the relevant websites of the Ministry of Education, the websites of individual schools, and the websites of the colleges within those schools. This data will be updated periodically according to the content updated on the websites. School and major-related data can be collected through web scraping or manual collection. The collected data will be initially processed into a hierarchical two-dimensional table and stored in the database for use in other steps.

[0058] 2) Preprocess the obtained school and major-related data to obtain the word sets corresponding to schools and majors; refer to Figure 3 For preprocessing the school and major-related data to obtain the word sets corresponding to schools and majors, the method is as follows:

[0059] 2.1) Construct a word segmentation model and load the stop word dictionary and word segmentation dictionary;

[0060] 2.2) Process each piece of the obtained school and major-related data through the word segmentation model to remove stop words; the stop words here include modal particles and / or meaningless words, such as modal particles like "de", "le", or words without specific meanings; taking "Mathematics and Statistics" as an example, after removing the stop word "and" through preprocessing, it becomes: "Mathematics Statistics";

[0061] 2.3) Segment each piece of data after removing stop words through the word segmentation model to form individual word sets. Each word set corresponds to a piece of data in the school and major-related data, and the word sets do not affect each other; taking the "Mathematics Statistics" generated in the above steps as an example: segment "Mathematics Statistics" to form the word set {Mathematics, Statistics}.

[0062] The implementation of the word segmentation model in step 2) can be achieved using the open-source project jieba for word segmentation, or other open-source projects, such as: IKAnalyzer, word2vec deep learning model, etc. This example does not make any restrictions; according to actual implementation experience, jieba word segmenter can be preferentially used to construct the word segmentation model. Its advantage is that it is more friendly and accurate for segmenting Chinese documents, and the intermediate data generated during processing and the result data generated after processing will be processed into a new two-dimensional table with a related hierarchy and stored in the database for subsequent steps to use, without overwriting the original data in the database;

[0063] 3) According to the word sets corresponding to schools and majors, obtain the professional similarity between majors and accordingly obtain the professional association relationship between majors; the professional similarity between majors includes the professional similarity of each national standard major compared with each school major, and the professional similarity between each school major;

[0064] Among them, the method for obtaining the professional similarity between majors and accordingly obtaining the professional association relationship between majors according to the word sets corresponding to schools and majors is specifically as follows:

[0065] 3.1) Based on the word sets corresponding to schools and majors, the name similarity, semantic similarity, and course similarity among majors are calculated respectively. In this embodiment, the word sets corresponding to schools and majors include the word sets corresponding to the national standard major set data and the word sets corresponding to the major set data of each school. Among them, the word set corresponding to each national standard major corresponds to the name of a national standard major (i.e., a major recorded in the national standard major classification), and the word set corresponding to each school major corresponds to the name of a school major (i.e., a major offered by a school). The name similarity, semantic similarity, and course similarity among majors include the name similarity and semantic similarity between each national standard major and each school major, as well as the name similarity, semantic similarity, and course similarity between each school major. The specific calculation method is as follows:

[0066] ① The calculation method for the similarity of names between each national standard major and each school's major, and between majors at different schools, is as follows:

[0067] First, the words in the word sets corresponding to the national standard professional set data and the professional set data of each school are merged, deduplicated, and sorted to form a professional corpus dictionary D1 corresponding to each professional, with a length denoted as N1. It should be noted that the final professional corpus dictionary D1 and its length N1 will vary depending on the collected datasets. This calculation of the professional corpus dictionary D1 can be performed during the preprocessing steps of the school and professional related data, such as... Figure 3 As shown;

[0068] Then, refer to Figure 4 Based on a professional corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the national standard major and each word set corresponding to the majors of various schools. One-hot encoding generates several vectors for a word set c, where each dimension component is 0. Each vector corresponds to a word in the word set c, and the dimension of these vectors is determined by the length of the professional corpus dictionary. In this embodiment, the dimension of these vectors is the length N1 of the professional corpus dictionary D1. Then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, resulting in the position vector corresponding to each word in the word set. For example, in the word set {mathematics, statistics}, the vector corresponding to the word "mathematics" is {0, 0, ... 1, 0, ... 0}, and the vector corresponding to the word "statistics" is {0, 1, ... 0, 0, ... 0}.

[0069] The position vectors corresponding to all words in each word set are merged in dimension, resulting in N1-dimensional vectors that contain the position information of all words in each word set corresponding to national standard majors and majors of various schools in the professional corpus dictionary. For example, the merged vector corresponding to the word set {mathematics, statistics} is {0, 1, ... 1, 0, ... 0}. In this embodiment, to reduce storage pressure and speed up calculation, the directly merged N1-dimensional vector is compressed into a dense vector, and then the compressed dense vector is used as the merged vector for subsequent steps. Example: The vector {0, 1, ... 1, 0, ... 0} of the word set {mathematics, statistics} after simple merging is compressed into a dense vector {N1#1, 10#1, 1}, that is, the merged vector corresponding to the word set {mathematics, statistics} is the dense vector {N1#1, 10#1, 1}. Here, N1 represents the length of the vector, "#1, 10" means that the 1st and 10th positions in the vector are not 0, and "#1, 1" means that the values ​​of the non-zero positions are 1 and 1 respectively.

[0070] The similarity algorithm is used to calculate the similarity S1 for each pair of merged vectors. This calculated similarity S1 is then used as the name similarity between each national standard major and each school major, or between majors within each school. Therefore, this method can be used to calculate the name similarity between a national standard major and a school major, or between two school majors, simply by calculating the similarity S1 using the corresponding word set. Each school major's word set corresponds to the name of that major. The calculation result is saved as a set r{major name 1, major name 2, name similarity S1}, for example: {Mathematics and Statistics, Applied Statistics, 0.78}. This embodiment does not limit the similarity calculation method; cosine similarity algorithm is preferred, but other similarity calculation methods such as Euclidean distance, Hamming distance, and Jaccard similarity coefficient can also be used.

[0071] ② The semantic similarity calculation methods for comparing each national standard specialty with each school specialty, and for comparing specialties among schools, are as follows:

[0072] The similarity of names between each national standard major and each school major, or between different school majors, is assessed. If the name similarity between each national standard major and each school major, or between different school majors, exceeds the name similarity threshold, the corresponding semantic similarity is directly set to the first preset value. Otherwise, a word similarity calculation method based on a thesaurus is used to calculate the semantic similarity for each word set corresponding to the national standard major and each word set corresponding to each school major. The calculated similarity is then used as the semantic similarity between each national standard major and each school major, as well as the semantic similarity between different school majors.

[0073] In this embodiment, the name similarity threshold is set to 0.97. When comparing, if the similarity S1>=0.97, that is, the name similarity between each major is greater than or equal to the name similarity threshold, since the name similarity is high enough, semantic similarity judgment is no longer performed. Instead, the semantic similarity S2 is directly set to the first set value (the first set value in this embodiment is 1.0), and the result set r obtained by word similarity calculation is expanded to {major name 1, major name 2, name similarity S1, semantic similarity S2}, where S2=1.0.

[0074] If the similarity S1 < 0.97, meaning the name similarity between different majors is less than the name similarity threshold, semantic similarity judgment is required due to the low name similarity. A word similarity calculation method based on a thesaurus is used for semantic similarity calculation, where the thesaurus is the "Extended Version of the Thesaurus from the Information Retrieval Research Lab of Harbin Institute of Technology". The final semantic similarity S2 is then obtained, referring to... Figure 5 The specific calculation steps are as follows:

[0075] i) Remove duplicates from the word sets c of two major names (one national standard major and one school major, or two different school majors), keep the remaining words, and combine them in pairs to generate a set A{a1, a2, ...}. For example, the word set {mathematics, statistics} and the word set {application, statistics}, after deduplication, become {mathematics} and {application}. After combining them in pairs, we get a1: {mathematics, application}.

[0076] ii) If there are no duplicate words after deduplication, set s2 to 1.0 directly; otherwise, proceed to step iii).

[0077] iii) Query the code of each word in the word forest. It has a 5-level structure. If any word is not in the word forest, set s2 to 0.1 directly. Otherwise, proceed to step iv).

[0078] iv) If the first layer structure is different, set s2 to 0.1 directly; otherwise, proceed to step v).

[0079] v) If the second layer structure is different, calculate the coding difference k and the total number of branches m2 in the second layer, and finally obtain S2 = 0.65*cos(m1*π / 180)*((m1-k+1) / m1); otherwise, proceed to step vi).

[0080] vi) If the third layer structure is different, calculate the coding difference k and the total number of branches m3 in the third layer, and finally obtain S2 = 0.80*cos(m2*π / 180)*((m2-k+1) / m2); otherwise, proceed to step vii).

[0081] vii) If the fourth layer structure is different, calculate the coding difference k and the total number of branches m4 in the fourth layer, and finally obtain S2 = 0.90*cos(m3*π / 180)*((m3-k+1) / m3); otherwise, proceed to step viiii).

[0082] viii) If the fifth layer structure is different, calculate the coding difference k and the total number of branches m5 in the fifth layer, and finally obtain S2 = 0.96*cos(m4*π / 180)*((m4-k+1) / m4); otherwise, S2 = 1.0.

[0083] The calculated semantic similarity S2 is merged into the set r, expanding it to {professional name 1, professional name 2, name similarity S1, semantic similarity S2}, where S2 is the calculation result of steps i) to viii). The parameters in the calculation formulas of steps i) to viii) are set according to the actual semantic situation to make the similarity calculation result as close to reality as possible. In other embodiments, it can be further optimized according to the actual similarity calculation result.

[0084] ③ Since the terminology set corresponding to schools and majors also includes the terminology set corresponding to the course data offered by each school and major; where the terminology set corresponding to the course data offered by each school and major corresponds to all courses offered under a given school and major name; then refer to Figure 6 The method for calculating the course similarity between different school majors is as follows:

[0085] The word sets corresponding to the course data of each school's major are merged, deduplicated, and sorted to form a course corpus dictionary D2 corresponding to the courses offered by each school's major, and its length is denoted as N2.

[0086] Based on the course corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the course data of each school major: For a word set, several vectors are generated, each with a component of 0 in each dimension. Each vector corresponds to a word in the word set, and the dimension of these vectors is determined by the length N2 of the course corpus dictionary. In this embodiment, the dimension of these vectors is the length N2 of the course corpus dictionary D2. Then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, resulting in the position vector corresponding to each word in the word set. For example, the courses offered by the Mathematics and Statistics major (i.e., the course data of the school major) are: Probability Theory, Data Analysis, Geometric Algebra, and Mathematical Statistics. Their positions in the professional course dictionary D2 are 1, 3, 5, 7, respectively. The combined vector is {1, 0, 1, 0, 1, 0, 1, 0, 0, 0, ...}, with a length of N2.

[0087] The position vectors corresponding to all words in each word set are merged in dimension to form vectors that contain the position information of all words in each word set in the course corpus dictionary. In this embodiment, in order to reduce storage pressure and speed up calculation, the N1-dimensional vector formed by direct merging is compressed into a dense vector, and then the compressed dense vector is used as the merged vector for subsequent steps. For example, the merged vector {1, 0, 1, 0, 1, 0, 1, 0, 0, 0, ...} is compressed into a dense vector {N2#1, 3, 5, 7#1, 3, 5, 7}, where N2 represents the length of the vector, #1, 3, 5, 7 represent that the 1st, 3rd, 5th and 7th positions in the vector are not 0, and #1, 3, 5, 7 represent that the values ​​of the non-zero positions are 1, 1, 1, 1 respectively.

[0088] The similarity algorithm is used to calculate the similarity between each merged vector pairwise. The calculated similarity S3 is taken as the course similarity between each major and expanded into the set r. The expanded set is: {Major Name 1, Major Name 2, Name Similarity S1, Semantic Similarity S2, Course Similarity S3}. It should be noted that when calculating the similarity between a country and each school's major, since there is no course-related data for the national standard major, the course similarity S3 does not exist. In this case, there is no need to expand the set r. That is, the set r used when calculating the major similarity between each national standard major and each school's major is still: {Major Name 1, Major Name 2, Name Similarity S1, Semantic Similarity S2}. This embodiment does not limit the similarity calculation method. The cosine similarity algorithm can be used first, or other similarity calculation methods such as Euclidean distance, Hamming distance, Jaccard similarity coefficient, etc. can also be used.

[0089] 3.2) Multiply the name similarity and semantic similarity of each national standard major with each school major by the corresponding weighting coefficients and then sum them to obtain the major similarity between each national standard major and each school major; multiply the name similarity S1, semantic similarity S2 and course similarity S3 of each school major by the corresponding weighting coefficients and then sum them to obtain the major similarity between each school major.

[0090] Reference Figure 7 The final weighted professional similarity S is calculated as follows: If set r contains name similarity S1, semantic similarity S2, and course similarity S3, the formula for calculating the final weighted professional similarity S is: S = S1 * 0.3 + S2 * 0.2 + S3 * 0.5, and set r is updated to {Professional Name 1, Professional Name 2, Professional Similarity S}; if set r only contains name similarity S1 and semantic similarity S2, the formula for calculating the final weighted professional similarity S is: S = S1 * 0.3 + S2 * 0.2 + S3 * 0.5. =S1 + S2 * 0.4, and update the set r to {major name 1, major name 2, major similarity S}; in other embodiments, the weighting coefficients of the two calculation formulas can be adjusted according to the actual situation; that is, in this embodiment, when calculating the similarity between national standard majors and school majors, S is calculated as: S = S1 * 0.6 + S2 * 0.4; when calculating the similarity between majors between schools, S is calculated as: S = S1 * 0.3 + S2 * 0.2 + S3 * 0.5.

[0091] Then, the similarity S between each national standard major and each school's major, as well as between majors within each school, is compared with the set similarity threshold and equality threshold. Where the similarity threshold is less than the equality threshold, the professional association relationships between each major are obtained accordingly.

[0092] If the similarity between two majors is greater than or equal to the similarity threshold but less than the equality threshold, then the professional association between these two majors is set as similar; if the similarity between two majors is greater than or equal to the equality threshold, then the professional association between these two majors is set as equal; if the similarity between two majors is less than the similarity threshold, then it is determined that there is no professional association between these two majors, and the corresponding data is discarded. In this embodiment, the similarity threshold is set to 0.7, and the equality threshold is set to 0.95; that is, if S < 0.7, then the set r{major name 1, major name 2, professional similarity S} corresponding to the professional similarity S is discarded; if 0.95 > S >= 0.7, then the association L1 is set as similar; if S >= 0.95, then the association L1 is set as equal; and accordingly, the set r{major name 1, major name 2, professional similarity S} that was not discarded is updated to the set r{major name 1, major name 2, professional association L1}.

[0093] 4) Based on the obtained school and major-related data, obtain the information relationships between each school and major; based on the professional relationships between majors obtained in step 3) and the information relationships between each school and major, generate a school professional knowledge map.

[0094] Since the acquired school and major-related data includes school and major-related data as well as basic data of each school and major and relevant information of each school; among which, the basic data of each school and major includes major ranking, major training direction and former name of major; and the relevant information of each school includes school ranking and school level data, the information correlation of each school and major itself includes basic information correlation of school and major L2, basic information correlation of school L3 and the correlation of courses offered by school and major L4.

[0095] like Figure 8 As shown, based on the acquired school and major-related data, the information association relationships of each school and major are obtained as follows: L2 is obtained through the basic data of each school and major; L3 is obtained through the relevant information of each school; L4 is obtained through the data of courses offered by each school and major; different entities are constructed based on the acquired school and major-related data; in this embodiment, the constructed entities include national standard major entities E1, school major entities E2, school name entities E3, school basic information entities E4, school major basic information entities E5, and major course offering entities E6.

[0096] The constructed entities E1-E5, along with the professional association L1 and various information associations L2-L4, are loaded into the graph database Neo4j. Matching and mapping are then performed to generate a school professional knowledge graph. The generated school professional knowledge graph is shown below. Figure 9 As shown, this information can be used by job seekers and employers to assess whether a candidate's major meets the job requirements.

[0097] Therefore, the school professional knowledge graph constructed by the method in this embodiment can help college students seeking employment and corporate recruiters understand the relationship between their majors and the required majors more clearly. It can also guide job seekers to apply for jobs with different major names but the same direction, such as accounting majors applying for jobs in financial management. At the same time, it can also help corporate recruiters understand the applicant's school information, major information, courses offered in the major, and whether they match the required abilities for the position more clearly. Furthermore, the constructed school professional knowledge graph does not only distinguish similarity based on the major name, but also combines multiple dimensions and methods to calculate a more accurate similarity, which can better solve the problems of talent omission and increased recruitment and employment costs encountered by college students in the process of employment and corporate recruiters in the recruitment process.

[0098] Example of a system for constructing a school knowledge graph:

[0099] This embodiment provides a technical solution for a school professional knowledge graph construction system; the school professional knowledge graph construction system includes a processor for executing program instructions to implement the school professional knowledge graph construction method as described in the above embodiment.

[0100] Since the specific working method and working principle of the school's professional knowledge graph construction system have been described in detail in the above-mentioned school professional knowledge graph construction method embodiment, they will not be repeated here.

[0101] This invention has the following characteristics:

[0102] 1) The constructed school professional knowledge map is not only distinguished by the similarity of the major name, but also calculates a more accurate similarity by combining multiple dimensions and methods. It can better solve the problems of talent omission and increased recruitment and employment costs encountered by college students in the process of employment and corporate recruiters in the recruitment process.

[0103] 2) The constructed school professional knowledge map not only includes professional relationships, enabling college students seeking employment and corporate recruiters to more clearly understand the relationship between their major and the required major, but also includes the information relationships of each school's majors themselves. It can guide job seekers to apply for jobs with different major names but the same direction, and can also help corporate recruiters to make a preliminary assessment of job seekers' actual abilities from the perspective of school majors, making it easier to judge whether job seekers are competent for the corresponding jobs.

[0104] It should be understood that the above-described specific embodiments of the present invention are merely illustrative or explanatory of the principles of the present invention, and do not constitute a limitation thereof.

Claims

1. A method for constructing a school knowledge map, characterized in that, The steps are as follows: 1) Obtain relevant data on schools and majors, including national standard major collection data, major collection data of each school, and course data of each school's majors; majors of each school include different majors offered by different schools and different majors offered by the same school. 2) Preprocess the data related to schools and majors to obtain the word sets corresponding to schools and majors, including the word sets corresponding to the national standard major set data and the word sets corresponding to the major sets of each school; each word set corresponding to a national standard major corresponds to a national standard major name, and each word set corresponding to a school major corresponds to a school major name; 3) Based on the set of words corresponding to schools and majors, obtain the name similarity and semantic similarity between each national standard major and each school major, and the name similarity, semantic similarity and course similarity between each school major, and then calculate the major similarity between each major and obtain the major association relationship between each major accordingly. The method for calculating name similarity is as follows: The words in the word sets corresponding to the national standard professional set data and the professional set data of each school are merged, deduplicated, and sorted to form a professional corpus dictionary corresponding to the majors; Based on the professional corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the national standard major and each word set corresponding to the majors of various schools: For a word set, several vectors are generated, each with a component of 0 in each dimension. Each vector corresponds to a word in the word set, and the dimension of these vectors is determined by the length of the professional corpus dictionary; then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, thus obtaining the position vector corresponding to each word in the word set; The position vectors corresponding to all words in each word set are merged in dimension to form vectors that contain the position information of all words in each word set corresponding to the national standard major and the majors of each school in the professional corpus dictionary. Then, a similarity algorithm is used to calculate the similarity of each merged vector pairwise. The calculated similarity is used as the name similarity between each national standard major and each school major or the name similarity between each school major. 4) Based on the acquired data related to schools and majors, obtain the information relationships between each school and major; based on the professional relationships between majors and the information relationships between each school and major, generate a professional knowledge map of the schools.

2. The method for constructing a school professional knowledge graph according to claim 1, characterized in that, The method for preprocessing school and major-related data to obtain the word set corresponding to schools and majors is as follows: A word segmentation model is constructed, a stop word dictionary and a word segmentation dictionary are loaded, and each piece of data in the acquired school and major-related data is processed by the word segmentation model to remove stop words; the stop words include modal particles and / or meaningless words; Each data point after removing stop words is segmented using a word segmentation model to form various word sets. Each word set corresponds to a data point related to a school and its major, and the word sets do not affect each other.

3. The method for constructing a school professional knowledge map according to claim 1, characterized in that, The similarity between different majors includes the similarity between each national standard major and each school's major, as well as the similarity between majors in different schools. The method for calculating the professional similarity between different majors is as follows: The similarity scores of the names and semantics of each national standard major compared with each school's major are multiplied by their respective weighting coefficients and then summed to obtain the major similarity scores of each national standard major compared with each school's major. The similarity scores of the names, semantics, and courses of each school's majors are multiplied by their respective weighting coefficients and then summed to obtain the major similarity scores of each school's majors.

4. The method for constructing a school professional knowledge map according to claim 3, characterized in that, The semantic similarity calculation methods for comparing national standard majors with individual school majors, and for comparing majors within different schools, are as follows: The similarity of names between each national standard major and each school major, or between different school majors, is assessed. If the name similarity between each national standard major and each school major, or between different school majors, is greater than or equal to the name similarity threshold, the corresponding semantic similarity is directly set to the first preset value. Otherwise, a word similarity calculation method based on a thesaurus is used to calculate the semantic similarity for each word set corresponding to the national standard major and each word set corresponding to each school major. The calculated similarity is used as the semantic similarity for each national standard major compared to each school major, and for each school major compared to each other.

5. The method for constructing a school professional knowledge map according to claim 3, characterized in that, The terminology set corresponding to schools and majors also includes the terminology set corresponding to the course data of each school's majors; among them, the terminology set corresponding to the course data of each school's majors corresponds to all courses offered under the name of a school's major. The method for calculating the course similarity between the various school majors is as follows: The word sets corresponding to the course data of each school's majors are merged, deduplicated, and sorted to form a course corpus dictionary corresponding to the courses offered by each school's majors. Based on the course corpus dictionary, one-hot encoding is performed on each word in each word set corresponding to the course data offered by each school major: several vectors with each dimension component set to 0 are generated for each word set, each vector corresponding to each word in the word set, and the dimension of these vectors is determined by the length of the course corpus dictionary; then, based on the position of each word in the word set in the corpus dictionary, the component in the vector corresponding to each word is marked as 1, thus obtaining the position vector corresponding to each word in the word set; The position vectors corresponding to all words in each word set are merged in dimension to form vectors that contain the position information of all words in each word set in the course corpus dictionary. Then, a similarity algorithm is used to calculate the similarity between each merged vector, and the calculated similarity is used as the course similarity between each major.

6. The method for constructing a school professional knowledge map according to any one of claims 1-5, characterized in that, The corresponding method for obtaining the professional relationships between various majors is as follows: The similarity between each national standard major and each school major, as well as between majors in each school, is compared with the set similarity threshold and equality threshold, where the similarity threshold is less than the equality threshold. If two majors have a similarity greater than or equal to the similarity threshold but less than the equality threshold, then the major association between the two majors is set to similar; if two majors have a similarity greater than or equal to the equality threshold, then the major association between the two majors is set to equal; if two majors have a similarity less than the similarity threshold, then it is determined that there is no major association between the two majors, and the corresponding data is discarded.

7. The method for constructing a school professional knowledge map according to any one of claims 1-5, characterized in that, The data related to schools and majors also includes basic data on each school's majors and relevant information about each school; the basic data on each school's majors includes major ranking, major training direction, and data on former names of the major; the relevant information about each school includes school ranking and school level data. The information relationships between each school's majors include the relationships between basic information about the school's majors, the relationships between basic information about the school, and the relationships between the courses offered by the school's majors. Based on the obtained data related to schools and majors, the information relationships between each school and its majors are obtained in the following ways: by obtaining the basic information relationships between schools and majors through the basic data of each school's majors; by obtaining the basic information relationships between schools through relevant information of each school; and by obtaining the information relationships between the courses offered by each school's majors through the data of the courses offered by each school's majors.

8. The method for constructing a school professional knowledge map according to any one of claims 1-5, characterized in that, Based on the aforementioned professional relationships and the information relationships within each school's respective majors, the method for generating a school's professional knowledge map is as follows: Based on the acquired school and major-related data, different entities are constructed; the constructed entities include national standard major entities, school major entities, school name entities, school basic information entities, school major basic information entities, and major-offered course entities. The constructed entities are matched with the professional relationships between various majors and the information relationships between the majors of each school, respectively, to generate a school knowledge knowledge graph.

9. A system for constructing a school knowledge map, characterized in that, Includes a processor for executing program instructions to implement the method for constructing a school knowledge graph as described in any one of claims 1-8.

Images